I've been debating this for a while with myself and I am leaning one way but was hoping for more input.
We have a HA FAS3070 netapp pair running ontap 7.3.7 with 4 attached disk shelves with 300 GB 15K FC disks.
The aggregates will host volumes that contain VM's mounted via NFS to the ESXi servers. Typically I would just create an aggregate per disk shelf. So say 4 aggregrates with 13 or 14 disks per. This way if we have client VM's that are going crazy and causing a tremendous amount of disk I/O it will only affect the VMs in the aggregrate and not the entire environment. However if we create a giant pool of disks using all of the disks available on all 4 disk shelves perhaps the impact of a couple VM's that are causing an abnormal amount of disk I/O won't affect much. I am building out a new environment which is why I am going back to this again. Any input is appreciated.